Note When a publication server acts as its own distribution server, the following information applies to the distribution database on the publisher. When a publication server uses a remote distribution server, the following information applies to the distribution database on that remote distribution server.
A distribution server retains individual transactions within the tables of the distribution database until they are applied to all eligible subscribers. The distribution server also preserves a transaction ID within the MSjobs table of its distribution database. This provides a pointer into the publication servers' transaction log, providing access to the last distributed transaction that was successfully transferred from that publication server into the distribution database.
There is no replication-specific information preserved within the transaction log of a distribution server.
For a distribution server to be automatically recovered using a previously performed database or transaction log backup, after recovery the MSjobs table must contain a valid pointer to a transaction that still resides within the publication server's transaction log. This pointer defines the relationship between these servers.
Transaction log dumps of published databases truncate the publishing database's transaction log and may remove the transaction pointed to by the image of the MSjobs table preserved in an earlier backup of the distribution server.
Important To automatically recover a distribution server, the database or transaction log dump used must recover the distribution server to a point in time later than the last transaction log dump of all associated publication servers. To ensure that the distribution server can always be automatically recovered from a backup, the administrator needs to establish a coordinated backup of these associated servers to ensure that a backup is performed on the distribution server after every transaction log backup of any associated publication server.